New thread for Engine Draw Claims

[hgm]

Fortunately we will not allow one person to dictate what others can and cannot do. We will not even allow an organization run by criminals to dictate what others can or cannot do. Tournament directors are free to choose which rules apply in their tournaments, and f*ck the FIDE. FIDE rules are made to deal with Humans, not with computers, so I see no reason whatsoever to even make an attempt to force an ill-suited set of rules through the trhroat of TDs and testers. People should be free to do what they want, no matter how much certain engine programmers dislike that!

------------------
The Free Chess Foundation

[bob]

As far as tablebase draws go, I am in 100% disagreement. What if only one of the engines is using tablebases? Let 'em play the game out. If both use tables, how long will the game last until the 50 move rule kicks in? 1 second total? Again, for FIDE rules, endgame tables have no meaning. Follow that here...

How long? It seems like quite a while for Crafty if swindle[*] mode is on. That is optional of course, but so would be the TB adjudication by the GUI. People who don't like it wouldn't be forced to use it.

*I am just assuming this is still supported.

If you play crafty against a program using endgame tables, why on earth would you have "swindle mode" enabled? That is designed for play against an opponent without endgame tables, to give them an opportunity to lose a drawn ending. Against tables it is pointless.

[bob]

Fortunately we will not allow one person to dictate what others can and cannot do. We will not even allow an organization run by criminals to dictate what others can or cannot do. Tournament directors are free to choose which rules apply in their tournaments, and f*ck the FIDE. FIDE rules are made to deal with Humans, not with computers, so I see no reason whatsoever to even make an attempt to force an ill-suited set of rules through the trhroat of TDs and testers. People should be free to do what they want, no matter how much certain engine programmers dislike that!

------------------
The Free Chess Foundation

and let anarchy reign supreme. Of course, perhaps the one trying to "define" the standards is the one that is out in left field???

[GothicChessInventor]

Adjudication without tablebases would not be fair because there are literally billions of endgames that cannot possibly be understood even with a long search.

Even "simple" endgames might confuse software without tablebases.



Consider the position above in Knight vs. 2 Pawns with white to move. What is the result that would be adjudicated?

It's actually a mate in 17 moves for the Knight!

Code: Select all

1. Kc1! c6 2. Ng2 Ka2 3. Kc2! Ka1 4. Ne3 Ka2 5. Nf5 c5 6. Nd6 Ka1 7. Nc4! Ka2 8. Nd2 Ka1 9. Kb3 c4+ 10. Kc3! {mutual zugzwang} Ka2 11. Kc2! Ka1 12. Ne4 {mutual zugzwang} Ka2 13. Nc3+ Ka1 14. Kc1 a2 15. Na4 c3 16. Nc5! c2 17. Nb3#

There are many examples that are even more troubling than this that are not so exotic either. I tried playing the winning side of the Q + P vs. Q endgame that was a mate in 268. I can't go 20 moves without giving up the draw even when I am on the side to deliver mate. There must be billions of draws in this tablebase.

[GothicChessInventor]

Fortunately we will not allow one person to dictate what others can and cannot do. We will not even allow an organization run by criminals to dictate what others can or cannot do. Tournament directors are free to choose which rules apply in their tournaments, and f*ck the FIDE. FIDE rules are made to deal with Humans, not with computers, so I see no reason whatsoever to even make an attempt to force an ill-suited set of rules through the trhroat of TDs and testers. People should be free to do what they want, no matter how much certain engine programmers dislike that!

------------------
The Free Chess Foundation

and let anarchy reign supreme. Of course, perhaps the one trying to "define" the standards is the one that is out in left field???

Actually, in terms of Gothic Chess computer tournaments, it has been proposed to waive the 50 move rule if:

A program announces a mate > 49 moves.
A program announces a mate > 19 moves and 20 moves or fewer remain for the 50 Move Rule draw.

To prevent abuse of this rule, such as a program announcing "mate in 300" on move 1, there is another rule:

After the game, if the program cannot demonstrate "mate in n-1" or quicker for every legal response to the initial mate announcment, nor play out the mate properly for both sides, then the opponent is awarded the draw.

[hgm]

Normally, such games would not be adjudicated at all, unless /adjudicateEndings is set to a value >= 5 and a tablebase for this particular ending would be available. Note that so far this is all science fiction, (for Gothic Chess even more so than for normal Chess, where bitbases with a compatible license already exist).

There is an adjudication option in WinBoard, where it listens to the consensus of the engines: if both engines agree that one of them leads by more than a user-set threshold for more than a user-specified number of moves, the game is adjudicated in favor of the leading engine. If both engines would recognize this as a badly lost position, e.g. through build-in recognizers, but would not have the tablebases to tell them what the winning move is, so that they still would need a long search to come up with a move, this type of adjudication could be used to shorten the game.

[hgm]

After the game, if the program cannot demonstrate "mate in n-1" or quicker for every legal response to the initial mate announcment, nor play out the mate properly for both sides, then the opponent is awarded the draw.

I must admit that I have some reservations w.r.t. this rule. It adds a lot of complexity, and I am not sure there is any practical benefit. Although I feel the same hurt as you when I see that a KBBKN win is killed by the 50-move rule because it happens to take 75 moves to force the conversion, such things happen extremely rarely. Many of the very long wins that are displayed on your website are found in end-games with Pawns, and would most likely not be draws even under the 50-move rule, as you would now and then advance the Pawn. You would need DTZ EGTBs to see that, and the long DTM does not necessarily tell you there is a problem. It could be that Gothic Chess in general has more lengthy end-games than normal Chess, because of the larger board, but if this is the case, the natural solution would be just to adopt, say, a 60-move rule in stead. We should really make an inventarization of DTZ requirements of the various end-games to see gow large the number of moves would have to be to solve 99.9% of all problems.

[GothicChessInventor]

After the game, if the program cannot demonstrate "mate in n-1" or quicker for every legal response to the initial mate announcment, nor play out the mate properly for both sides, then the opponent is awarded the draw.

I must admit that I have some reservations w.r.t. this rule. It adds a lot of complexity, and I am not sure there is any practical benefit. Although I feel the same hurt as you when I see that a KBBKN win is killed by the 50-move rule because it happens to take 75 moves to force the conversion, such things happen extremely rarely. Many of the very long wins that are displayed on your website are found in end-games with Pawns, and would most likely not be draws even under the 50-move rule, as you would now and then advance the Pawn. You would need DTZ EGTBs to see that, and the long DTM does not necessarily tell you there is a problem. It could be that Gothic Chess in general has more lengthy end-games than normal Chess, because of the larger board, but if this is the case, the natural solution would be just to adopt, say, a 60-move rule in stead. We should really make an inventarization of DTZ requirements of the various end-games to see gow large the number of moves would have to be to solve 99.9% of all problems.

The 50 Move Rule should only be for human players. Players get tired, and frustrated. Recall FIDE changed this rule a few times when Ken Thompson published his long wins in the 1980s.

Computers don't need the 50, 60, or even 200 move rule. What is needed is some way to keep human programmers from abusing the rule we come up with.

It there's a cool mate in 200+ moves, I'd like to see programs play it out!

[GothicChessInventor]

I have been thinking about releasing the tablebases for Gothic Chess, but it would be so much easier to just release the generator.

Just to refamiliarize myself with how long it took to do the 5-piece set, I am regenerating all of them. I started on Christmas Eve 2007 and I have 101 of the 196 5-piece endgames solved.

Right now Archbishop + Rook vs. Chancellor is being solved.

The remaining endgames will solve much faster than the earlier endgames since the strong side gets progressively stronger pieces and symmetry reduction for like material on the strong side greatly reduces the iteration time.

I forgot how much overhead tablebase verification adds to it all. Even with an average verification taking 90 minutes, for 196 endgames, it's nearly 2 weeks for this alone!

It looks like about 41 days will be required to solve the 5-piece set for Gothic Chess. I don't suppose many programmers will have the patience to endure this. Then, there's the adding of the access code, and the buffering code if you want to probe them in RAM like Gothic Vortex does.

[bob]

Fortunately we will not allow one person to dictate what others can and cannot do. We will not even allow an organization run by criminals to dictate what others can or cannot do. Tournament directors are free to choose which rules apply in their tournaments, and f*ck the FIDE. FIDE rules are made to deal with Humans, not with computers, so I see no reason whatsoever to even make an attempt to force an ill-suited set of rules through the trhroat of TDs and testers. People should be free to do what they want, no matter how much certain engine programmers dislike that!

------------------
The Free Chess Foundation

and let anarchy reign supreme. Of course, perhaps the one trying to "define" the standards is the one that is out in left field???

Actually, in terms of Gothic Chess computer tournaments, it has been proposed to waive the 50 move rule if:

A program announces a mate > 49 moves.
A program announces a mate > 19 moves and 20 moves or fewer remain for the 50 Move Rule draw.

To prevent abuse of this rule, such as a program announcing "mate in 300" on move 1, there is another rule:

After the game, if the program cannot demonstrate "mate in n-1" or quicker for every legal response to the initial mate announcment, nor play out the mate properly for both sides, then the opponent is awarded the draw.

I am again reminded of the question, "what is wrong with following the FIDE rules here, since we all aspire to beating humans in tournaments, rather than beating other computers. My efforts are _always_ geared toward playing humans. And there, we have a pretty precise set of rules we have to adhere to.